colony

Coordinate repository work through Colony MCP. Use when an agent starts, resumes, chooses work, claims files, hands off work, records working state, or finishes a task in a repo where Colony tools are available.

Bucket Claude Code and Codex session tokens into Colony tasks and surface improvement suggestions. Use when the user runs `/co:tokens`, asks how many tokens a task spent, why a session got expensive, or what to change next time.

This skill should be used when building data processing pipelines with CocoIndex, a Python library for incremental data transformation. Use when the task involves processing files/data into databases, creating vector embeddings, building knowledge graphs, ETL workflows, or any data pipeline requiring automatic change detection and incremental updates. CocoIndex is Python-native (supports any Python types), has no DSL, and uses version 1.0.0 or later.

Validate, three-way-merge, and archive a colonykit change. Use when the user runs `/co:archive <slug>` or says the change is done. Atomic — either the archive and root-write both land, or neither does.

Execute the next §T task on the current colonykit change. Use when the user runs `/co:build` or asks you to proceed with the next task. Pulls cite-scoped context, executes, on test failure records via backprop gating.

Open a new colonykit spec change. Use when the user runs `/co:change <slug>` or asks to start a new proposal, delta, or change-in-flight. Creates openspec/changes/<slug>/CHANGE.md and opens a task-thread on spec/<slug>.

Read-only drift check across root SPEC.md, the active change's delta, and the code. Use when the user runs `/co:check` or asks whether the spec and code are in sync. Writes nothing.

Use when reviewing active multi-agent work, PRs, or handoffs where you need Cavemem's hivemind tool to map branch and task ownership before reading full session memory.